Project Summary Patients undergoing radiation therapy (RT) for oral and craniofacial cancers such as human papillomavirus- positive oropharyngeal cancer (HPV+ OPC) experience a host of side effects caused by radiation-induced injury to healthy tissues. Although RT is highly curative for HPV+ OPC, radiation-induced sequelae can persist for decades of survivorship, significantly degrading a patient's oral health and quality of life. Toxicity to healthy tissues can be reduced by adaptive replanning, in which the geometry of the radiation beams is re-optimized periodically during a multi-week course of RT to account for tumor shrinkage and normal tissue deformation. Adaptive replanning is now clinically feasible for oral and craniofacial cancers due the recent development of a hybrid MRI/linear accelerator device (MR-Linac). Adaptive treatments have used only basic anatomical MRI pulse sequences to monitor the tumor volume. However, we propose an adaptive treatment strategy that uses a functional MRI technique called diffusion-weighted imaging (DWI), which can assess normal tissue function, identify radioresistant sub-volumes within tumors, and predict patient response to RT. The hypothesis of this study is that the functional information from DWI can be implemented into the adaptive replanning process for oral and craniofacial cancers such that it is clinically feasible (with the new MR-Linac device) and will reduce side effects. To test this hypothesis, we will first develop a multivariate regression model relating changes in ADC values of the tumor and healthy tissues to HPV+ OPC patient outcomes. This information will be integrated into a DWI-based adaptive replanning workflow for the MR-Linac (Specific Aim 1). Next, the DWI- based adaptive replanning approach will be modeled retrospectively on daily patient images. A dose accumulation algorithm compatible with the MR-Linac's MRI-based dose calculation method will be developed and employed to measure cumulative doses to organs at risk. Cumulative doses will be related to normal tissue complication probability models to determine whether this approach lowers the risk of side effects (Specific Aim 2). The expected outcome of these specific aims is the development of an adaptive RT approach that uses functional data from DWI. The clinical feasibility and benefit of this treatment scheme will be demonstrated through in silico and statistical modeling so that it may eventually be used in the clinic. This project will positively impact patients with HPV+ OPC by enabling the delivery of personalized, targeted RT to the tumor while sparing normal tissues and reducing side effects. Further, this work will have a broader impact on the field of oral, dental, and craniofacial health by introducing a novel treatment paradigm that directly monitors and reacts to normal tissue injury without compromising tumor control.